AI carbon footprint in Asia Pacific - Singapore: emissions per token and per query
Factor version soma-ai-ef-2026.10-dataset-v3.0.0. Last updated . Location-based grid averages, not renewable contracts. Estimates from published benchmarks.
Part of the SOMA AI Emission Factors cluster.
How much CO₂ does an AI query emit when served from Asia Pacific - Singapore?
A 1210-token query to a class B model (gpt-4o) served from Asia Pacific - Singapore emits about 0.055 g CO₂e, and a million tokens emit 0.0401 kg CO₂e.
Rank 16 of 16 for class B, lowest carbon first (full table). 5.6x the level of EU North - Sweden.
Which grid supplies the electricity in Asia Pacific - Singapore?
Asia Pacific - Singapore maps to Singapore, a national grid, at 500.9 g CO₂e per kWh for 2023. That is 2.1 times the EU-27 average (236 g) and 1.0 times the global average (483.2 g). The source is Ember yearly electricity data, CO2 intensity 2023, 500.87 g/kWh (lifecycle, IPCC AR5).
Cloud regions on this grid, 3 in all (1 AWS, 1 Azure, 1 Google Cloud).
| Provider | Region id | Region name | Location | Mapping confidence | Mapping note |
|---|---|---|---|---|---|
| AWS | ap-southeast-1 | Asia Pacific (Singapore) | Singapore | high | |
| Azure | southeastasia | Southeast Asia | Singapore | high | |
| Google Cloud | asia-southeast1 | asia-southeast1 | Jurong West, Singapore | high |
Which grids have an intensity close to Singapore?
| Difference g | Grid | Code | g CO₂e per kWh | Ratio |
|---|---|---|---|---|
| 0.3 | Bolivia | BOL | 501.2 | 1.001 |
| -0.9 | Kiribati | KIR | 500 | 0.998 |
| -0.9 | Vanuatu | VUT | 500 | 0.998 |
| -3.7 | RFC Michigan | RFCM | 497.2 | 0.993 |
| -3.9 | Guadeloupe | GLP | 497 | 0.992 |
| 5.0 | Haiti | HTI | 505.9 | 1.010 |
| -6.0 | Turkey | TUR | 494.9 | 0.988 |
| -7.3 | Malta | MLT | 493.6 | 0.985 |
| -8.3 | Japan | JPN | 492.6 | 0.983 |
| 9.4 | Nigeria | NGA | 510.3 | 1.019 |
Emission factors for Asia Pacific - Singapore by model class
| Model class | Electricity kg | Hardware kg | Construction kg | Training kg | Total kg CO₂e | g per query | Water L | mL per query |
|---|---|---|---|---|---|---|---|---|
| Class A (small) | 0.01090 | 0.00027 | 0.00014 | 0.00060 | 0.012 | 0.016 | not available | not available |
| Class B (mid-size) | 0.03571 | 0.00101 | 0.00046 | 0.00293 | 0.040 | 0.055 | not available | not available |
| Class C (frontier) | 0.13296 | 0.00388 | 0.00169 | 0.01284 | 0.151 | 0.21 | not available | not available |
| Model class | Central | Class range low | Class range high | High case | B200 | Development |
|---|---|---|---|---|---|---|
| Class A | 0.01191 | 0.007346 | 0.01728 | 0.01914 | 0.01043 | 0.01487 |
| Class B | 0.04011 | 0.0258 | 0.04873 | 0.06381 | 0.02802 | 0.04869 |
| Class C | 0.1514 | 0.05907 | 0.227 | 0.2397 | 0.1557 | 0.1813 |
How much does siting matter for a model served from Asia Pacific - Singapore?
Electricity is 89% of the class B total here (class A 91.5%, class C 87.8%), with hardware at 2.5%, construction 1.1% and training 7.3%. Electricity carries most of the 0.0401 kg, so the grid is the biggest lever.
The example company (2,323 million class B tokens a year) emits 93 kg CO₂e from Asia Pacific - Singapore. Moving to Asia Pacific - Japan would save 1.4 kg (1.5%) or US Central - Iowa would save 4.8 kg (5.1%).
| Region | Grid g per kWh | Class A | Class B | Class C |
|---|---|---|---|---|
| Asia Pacific - Japan | 492.6 | 0.418 | 1.37 | 5.1 |
| US Central - Iowa | 472.1 | 1.46 | 4.77 | 17.8 |
How much water does AI usage consume in Asia Pacific - Singapore?
Water: not available for Asia Pacific - Singapore. The dataset publishes no off-site water factor for this region, so SOMA prints no water figure rather than a partial one.
Where do these numbers come from and how do I cite them?
Figures per query are one plain exchange, before the Tier 2b reasoning uplift. Electricity term. GPU energy per token from the ML.ENERGY Leaderboard v3, a facility overhead of 1.20, and grid carbon intensity from EPA eGRID 2023 and Ember 2023. Embodied hardware, data-centre construction and amortised training terms are documented on the methodology page. Water combines on-site cooling water (Li et al. 2025) with the water embedded in grid electricity (Reig et al. 2020). The modelled range of the models inside a class, around the class value, is −41% to +48% for class A, −39% to +24% for class B and −67% to +55% for class C.
Dataset: AI Inference Emission and Resource Factors for Corporate GHG Inventories (SOMA AI Emission Factors (SAEF)), Llopis-Garcia, G., SOMA AI, 19 October 2026, CC BY 4.0, concept DOI 10.5281/zenodo.20443585, which resolves to the latest version (this version 10.5281/zenodo.23371447, earlier version 2 values at 10.5281/zenodo.22767475). Method: Accounting for Purchased AI Inference in Corporate GHG Inventories: A Token-Based Method, arXiv:2606.10660, under peer review.
Partners can fetch any row from the AI factors API, and the whole table is a CSV download.
Cite as "0.0401 kg CO₂e per million tokens, class B, Asia Pacific - Singapore, SAEF, 19 October 2026" and link somaai.earth/ai-footprint/factors/singapore.